The mechanisms used by the brain to adapt its function in relation to time are still poorly understood. Scientists from the Institut Pasteur used an artificial neural network to model the activity of the cerebellum, the brain region responsible for motor coordination and timing of cognitive operations.
The brain is undoubtedly one of the most mysterious organs of the body (see the report Unraveling the brain's mysteries). It is not just capable of receiving information; it can also date that information chronologically to the hour. This is how circadian rhythms, which enable us to alternate between wakefulness and sleep depending on the time of day, are established. But we sometimes have to process information in less than a second, for example when making quick decisions or performing complicated movements. This is the role of the cerebellum, a brain region located at the back of our head, above the brainstem. Certain neurons in the cerebellum are activated in these situations, but it was not previously known whether they act separately or together to measure time.
Scientists in the Synapse and Circuit Dynamics Unit at the Institut Pasteur in Paris used an artificial neural network similar to those used in deep learning. "This model reproduces the neural activity of animals during specific behaviors, such as the reflex of closing an eye just before an unpleasant stimulus," describes David DiGregorio, unit head and co-author of the study.
As messages are passed on via successive neurons, they are compared and compiled to form a single message. This means that when faced with a stimulus, a response of variable length is emitted, and is potentially integrated by learning to generate precise movements. Credit: Barri and DiGregorio (Synapse and Circuit Dynamics)
"Because the network contains numerous dynamic synapses, it can generate electrical impulses for different temporal frameworks. All these signals are grouped together to form a basis for generating new, precisely timed impulses that make up the final message," explains Alessandro Barri, co-author of the study. Reproducing this pattern over time may lead to the development of a faster and even more specific response to the stimulus, exactly like during the learning process.
Synaptic basis of a sub-second representation of time in a neural circuit model. Nature Communications, 22 décembre 2022. https://doi.org/10.1038/s41467-022-35395-y
Barri, A.1, Wiechert, M.T.1, Jazayeri, M.2 et D. A. DiGregorio1.
1. Institut Pasteur, Université Paris Cité, Synapse and Circuit Dynamics Laboratory, CNRS UMR 3571, Paris, France
2. McGovern Institute for Brain Research, Massachusetts Institute of Technology, Cambridge, MA, USA
Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA